Traditional methods of cooling electronic systems, such as computers, have most commonly involved air as the heat transfer medium within the system. In such systems, heat generated by electronic components is typically transferred to the air surrounding the components and then to an ultimate sink, generally the room in which the electronic system is located. Heat transfer between the components and the surrounding air may be enhanced by increasing the surface area of contact between the electronic components and the air, for example, by using a heat sink, which itself may have fins to further increase the surface area available for heat transfer. Air circulation within the system may occur by natural convection or may be further enhanced by forcing the air to circulate about the components by means of a fan or blower.
As electronic equipment has become more sophisticated and yet more compact in size, the density of the heat-dissipating components mounted on a particular circuit board has necessarily increased. This is also true of the integrated circuit packages themselves, with the number and density of active devices, such as transistors, within a given package steadily increasing over time.
Unfortunately, these trends of increasing complexity and decreasing size have meant that the amount of heat dissipated within a given volume has also increased. With current-day integrated circuit technologies, the time is fast approaching when conventional air convection cooling methods, even those using forced air circulation with large heat sinks, will not adequately maintain certain high density integrated circuit packages within their permissible operating temperature range.
Consider, for example, the Pentium.TM. series of microprocessor chips recently introduced by Intel Corporation of Santa Clara, Calif. Depending on its operating speed, a Pentium.TM. P-5 chip typically dissipates on the order of 15 watts in a package which has less than two square inches of surface area, and a Pentium.TM. P-6 chip dissipates from 22 to 30 watts. Other microprocessor chips, such as the Alpha.TM. microprocessor recently introduced by Digital Equipment Corporation of Maynard, Mass. are projected to produce heat dissipations of 25 watts or higher in the same two square inches in their highest speed versions. If conventional air convection cooling techniques are to be used, such chips would require a very high air flow rates over finned heat: sinks in order to maintain the chips in their desired operating temperature range.
Although such forced air cooling requirements are technically feasible, practical considerations rule out their use in current day personal computer applications. One reason is that the blowers or fans necessary to generate such a high air flow rate would necessarily create an unacceptable noise level in an operating environment such as an office which is expected to remain relatively quiet. This problem is further exacerbated in applications such as laptop and notebook computers, where the additional noise and weight of a forced air cooling system is simply neither practical nor desired.
Consequently, there has been renewed interest in adapting liquid cooling techniques to the problem of cooling high-powered integrated circuits. These techniques generally fall into two broad groups, single-phase and two-phase cooling systems. In a single-phase liquid cooling system, the coolant remains in the liquid phase over the normal and expected system operating temperature range. In a two-phase system, the coolant changes from the liquid phase to the vapor phase during at least one point in the normal operating temperature range.
One example of a single-phase liquid cooling system makes use of a hermetic enclosure filled with a high-boiling point liquid coolant which completely encloses the heat generating component. The enclosure may also be provided with external fins. Heat is transferred from the heat dissipating component to the liquid coolant by conduction and from the coolant to the walls of the enclosure by natural convection. The enclosure itself may further be cooled by circulating air around it. Such a cooling method can be effective, but involves other problems such as chemical incompatibilities between the component and the coolant over the long-term, and the difficulty of obtaining access to the component for maintenance.
Other single-phase systems do not directly immerse the integrated circuit component in the liquid, but instead confine just the coolant to a container which is then placed in intimate contact with the component. Heat is thus conducted from the component through the container wall into the liquid, which then dissipates the heat by natural convection.
One embodiment of the latter single-phase system uses a container in the form of a sealed flexible bag which is completely filled with a liquid coolant. The bag is typically constructed from a flexible plastic film which is relatively impermeable to both the air and the enclosed liquid. Metal inserts or thermal vias, which pass through the wall of the bag, may also be used with this type of system to more efficiently conduct the heat from the component to the coolant. Examples of such coolant bag systems are shown in U.S. Pat. Nos. 4,997,032 and 5,000,256.
While such single-phase bag systems can be useful in certain situations, they have several disadvantages. Because they use a single-phase coolant, the available heat transfer rate is still relatively low. Consequently, they cannot typically be used with the high heat dissipating electronic components such as microprocessors. In addition, the bags have a relatively large volume which conflicts with the current trend of reducing system size as much as possible, and thus single-phase bag systems have not found widespread practical application.
Two-phase liquid cooling systems have increasingly been used to overcome the limitations associated with single-phase systems. In a two-phase system, as the component heats up, a liquid coolant is vaporized. The vapor then travels to a condenser section of the system, where the coolant vapor is converted back into a liquid. The liquid is returned by some means to the heat dissituating component and the boiling and condensing cycle is continuous.
Such a two-phase device is shown in U.S. Pat. No. 3,741,292. In that system, the heat dissipating component is placed within a hermetic enclosure and directly immersed in a pool of low boiling point, dielectric liquid coolant. The heat dissipated by the component causes the liquid to boil, and the resulting vapor is collected in an enclosure space located above the liquid pool. The enclosure space is filled with inwardly extending fins which serve as a condenser for the coolant vapor. As the vapor condenses, it runs back into the liquid pool under the influence of gravity.
Other two-phase cooling systems, so called heat pipe systems, do not directly immerse the component in the coolant. Such a system consists of an elongated hermetic container having walls made with a thermally conductive material such as copper. One end of the container acts as an evaporator and the other end acts as a condenser. A wick or other capillary device, such as a fine mesh screen, extends along the interior of the container. During manufacture of the heat pipe, the container is partially filled with liquid coolant, and an) residual, non condensing gases, such as air, are purged, and the container is then sealed. The heat dissipating component is mounted to the evaporator end of the pipe, and heat is transferred by conduction through the container wall. As the coolant evaporates, or boils, the resulting vapor travels through the container to the other end where it condenses back to a liquid. The liquid is then returned to the evaporator end by means of the wick.
Although the direct immersion and heat pipe techniques can transfer heat away from the heat dissipating component quite efficiently, they also have their limitations. More specifically, both techniques use rigid, hermetically sealed containers. When the ambient temperature changes, the pressure inside the container changes, with a consequent change in the boiling point of the coolant. In addition, because the container is evacuated, there exists a significant pressure differential across the walls of the container. As the container is exposed to repeated heating and cooling cycles, the repeated change in pressure differential causes the walls of the container to flex. Eventually, the container fatigues, causing small leaks. When a leak does occur, air is drawn into the container. Later, when the component is then reactivated, the presence of air increases the pressure inside the container and may cause some of the liquid to be driven out of the container, thereby compromising the cooling capability of the device. Consequently, such devices are typically not considered to be useful in environments where long-term, low-maintenance operation is required.
Furthermore, the increase in internal pressure results in an increase in the coolant boiling point, which may also be altered by the presence of any residual air introduced into the system by leaks. Such devices thus cannot be expected to have a single predictable boiling temperature.
Prior art two-phase systems are also prone to a phenomenon called thermal hysteresis, or "overshoot". This occurs during device warmup as a result of the fact that the coolant may not begin to boil when the device temperature initially reaches the nominal boiling point. Instead, the tendency is for the temperature to continue to increase past the boiling point, and then for boiling to suddenly erupt. Once coolant boiling finally does occur, the device temperature returns to its normal operating range. However, in the interim, the system has been temporarily subjected to a temperature well above the boiling point. Overshoot is a highly undesirable condition as it stresses the components being cooled and in some cases may even cause them to temporarily operate; outside their expected temperature range.
A detailed analysis of thermal hysteresis is provided by Bar-Cohen in Hysteresis Phenomena at the Onset of Nucleate Boiling. The full extent of the phenomenon is discussed along with some approaches shown to alleviate the problem.